Activation of Autophagy in a Rat Model of Retinal Ischemia following High Intraocular Pressure

Acute primary open angle glaucoma is an optic neuropathy characterized by the elevation of intraocular pressure, which causes retinal ischemia and neuronal death. Rat ischemia/reperfusion enhances endocytosis of both horseradish peroxidase (HRP) or fluorescent dextran into ganglion cell layer (GCL) neurons 24 h after the insult. We investigated the activation of autophagy in GCL-neurons following ischemia/reperfusion, using acid phosphatase (AP) histochemistry and immunofluorescence against LC3 and LAMP1. Retinal I/R lead to the appearance of AP-positive granules and LAMP1-positive vesicles 12 and 24 h after the insult, and LC3 labelling at 24 h, and induced a consistent retinal neuron death. At 48 h the retina was negative for autophagic markers. In addition, Western Blot analysis revealed an increase of LC3 levels after damage: the increase in the conjugated, LC3-II isoform is suggestive of autophagic activity. Inhibition of autophagy by 3-methyladenine partially prevented death of neurons and reduces apoptotic markers, 24 h post-lesion. The number of neurons in the GCL decreased significantly following I/R (I/R 12.21±1.13 vs controls 19.23±1.12 cells/500 µm); this decrease was partially prevented by 3-methyladenine (17.08±1.42 cells/500 µm), which potently inhibits maturation of autophagosomes. Treatment also prevented the increase in glial fibrillary acid protein immunoreactivity elicited by I/R. Therefore, targeting autophagy could represent a novel and promising treatment for glaucoma and retinal ischemia.


Introduction
Retinal ischemia, a common cause of blindness worldwide, involves reduced blood flow and impaired diffusion of oxygen; it is associated with acute and chronic glaucoma (primary open angle glaucoma, POAG), central or branch retinal arterial occlusion (CRAO/BRAO), retinal detachment, and diabetic retinopathy [1], [2], [3], [4]. In rats, ischemia associated with high intraocular pressure (IOP) [5], [6] produces pathological features that are almost identical to those reported for CRAO and POAG in humans.
Ischemia/reperfusion (I/R) injury is characterized by retinal degeneration, including extensive loss of neurons in the ganglion cell layer (GCL) and in the inner nuclear layer (INL) [7] the extent of the insult and the severity of neuronal death are related to the duration of ischemia, or the degree of IOP elevation [8], [9], [10]. Three modes of cell death, apoptosis, necrosis and autophagic death, have been described [11], [12], [13]: the first two, which can be readily identified, have been extensively investigated, whereas autophagic death has only recently garnered attention as a significant contributor to ischemia associated damage.
Three main forms of autophagy, chaperone-mediated, microand macroautophagy (hereafter simply called autophagy) have been described [14]. In eukaryotes, autophagy is a physiological process that leads to the degradation of long-lived proteins, cytoplasmic organelles and toxic agents by degradation in preexisting lysosomes [15]. Lysosomes, which contain different acid hydrolases, fuse with the new autophagic vacuole and load degradative enzymes into it [16]. Autophagy-associated cell death is caspase-independent, necrosis like [17], [11], [18] and apparently operates as an alternative mechanism when apoptosis has been compromised [19]. However, recent findings demonstrate the strong correlation with caspases [20]. Autophagic death is detected during development and tissue remodelling [21], [22], subsequent to ischemia/hypoxia [23], [24] and in a number of neurodegenerative diseases [25] in the retina, autophagy has been observed during development [26], in response to light exposure [27].
In this paper we document the occurrence of autophagic retinal cell death following I/R produced by acute IOP increase. We showed earlier that this model of I/R induces apoptotic cell death [10]; we now extend these observations to demonstrate that I/R also induces (i) autophagic activity, (ii) the formation of lysosomal vacuoles (detected by their content of acid phosphatase, AP), and promotes (iii) enhanced endocytosis, a process characteristic of dying neurons [28], [29], [30].
Taken together these results demonstrate enhanced autophagic flux. Moreover, our studies underline the important relationship between autophagy and apoptosis in the control of cell death after I/R, bearing implications for the development of potential neuroprotective therapies that are aimed at preventing ischemiarelated cell death.

Acid phosphatase histochemistry
Overall, retinal morphology was conserved following I/R ( Figure 1). In I/R retinas, AP activity was detected at 12 h following I/R (Figure 1 A-B), was maximal by 24 h (Figure 1 C-D) and declined at 48 h (Figure 1 E). Both methods used to visualize enzyme activity showed robust staining at 24 h postinsult, although staining was darker with the Barka and Anderson technique. Most intense staining was localized to GCL; sporadic positively stained cells were also visible in the inner nuclear layer (Figure 1 C-D, arrowheads). At high magnification, Gomori staining revealed clusters of small, intensely stained granules, preferentially located in the periphery of the cytoplasm (Figure 1 G), as is characteristic of lysosomal systems. Almost all GCLneurons were stained, but to different degrees: the larger the cell the more intensely reactive it was. The use of NaF in the incubation medium resulted in a complete inhibition of enzymatic activity. In control sections non-specific reactivity for AP was detectable in retinal neurons (Figure 1 F).

Labelling endocytosis in vivo
In response to I/R, endocytotic activity in retinal neurons was reflected in a strong, selective uptake of horseradish peroxidase (HRP) or of fluorescein isothiocynate (FITC)-labelled dextran. Twenty-four hours following I/R, HRP-or FITC dextran-positive granules were visible in ganglion cell layer, (Figure 1 H and I, respectively): labelling extended throughout the cells, and was particularly robust following HRP uptake.

Lysosomal and autophagosomal activity
Twelve and twenty-four hours after the I/R, lysosomeassociated membrane protein-1 (LAMP1) immunostaining was detectable in the GCL ( In fact, in the early stage, the initial vesicle, i.e. the phagophore, has formed and subsequently the LC3 complex associates to the developing double autophagosome membrane, demonstrating the autophagosome formation [31]. We investigated the relationship between autophagic and lysosomal activity using double-immunolabeling for LC3 and LAMP1 at 24 hours. Most LC3-positive neurons displayed also an increase in LAMP1 labelling, underlining that the two activities occurred in the same cells (Figure 2 G). At high magnification, fusion of autophagosomes with lysosomes could be appreciated (Figure 2 H, arrowheads).

Expression of LC3 in IOP retina after 24 h
Autophagy induction was determined by using the expression levels of the autophagy protein LC3. To evaluate the increases in autophagy flux after IOP we probed the retinal lysates with anti-LC3 antibody. The western blot analysis demonstrates that the antibody recognizes two LC3 isoforms, 18 (LC3-I) and 16 kDa (LC3-II). In IOP retinas LC3-II was upregulated approximately 20% compared with the sham (Figure 3). The intensities of the signals of LC3-I and LC3-II were normalized with tubulin.

Relationship between autophagic and apoptotic death
To evaluate the relationship between autophagic and apoptotic mechanisms, we investigated the expression of cleaved caspase-3, a critical executioner of apoptosis; it is partially or completely responsible for the proteolytic cleavage of many key proteins [32]. However, the association of several markers is required for appropriate detection of apoptotic cells [33]. Therefore, the identification of cleavated caspase-3 positive neurons should be related to other apoptotic markers, such as Terminal deoxynucleotidyltransferase-mediated biotinylated UTP Nick End Labeling (TUNEL)-staining.
At 24 hours post-I/R, LC3 and cleaved caspase-3 double labeled neurons were detected ( Figure 4, arrow in A and high magnification in B). Nevertheless, we could also find single labeled neurons for each marker, thus suggesting that autophagy and apoptosis do not necessarily overlap and may occur independently or at different time points from each other ( Figure 4, arrowheads in A). I/R in retina increases both autophagic [34] and apoptotic [10], [35]

Inhibition of autophagy prevents reactive astrogliosis in the retina
We also studied the effects of 3-MA on glial fibrillary acidic protein (GFAP) expression, the main intermediate filament specific for mature astrocytes in the central nervous system in normal and in pathological conditions [36]. In fact, in the control retina, GFAP was located exclusively in the end feet of the Müller cells ( Figure 5E

Effects of I/R +/2 3-MA treatment on the number of GCL-neurons
Counts of GCL-neurons following I/R showed a significant decrease in GCL-neurons number in rats treated with the vehicle alone (12.2161.13 cells/500 mm), compared to controls (19.2361.12 cells/500 mm); this decrease was partially prevented by 3-MA (17.0861.42 cells/500 mm, *P,0.05) ( Figure 6).

Discussion
The present study investigated the involvement of autophagy in a rat model of ischemia/reperfusion after elevated IOP. Increased IOP leads to a significant amount of apoptosis in the rat retina [10], as indicated by the activation of caspase-3-mediated mechanisms and by the presence of TUNEL-stained neurons. In addition, retinal ischemia also causes necrotic cell death [37]. Here we show that I/ R leads to the appearance of AP-positive granules, to the increase in LC3-II and LAMP1 expression and to enhanced endocytosis of both HRP and FITC labelled dextran in GCL-neurons.
In our experiments, AP-positive granules, characteristic of lysosomes, were present 12 and 24 h after the insult in GCLneurons. An increase in lysosomal profiles has also been observed ultrastructurally in the ischemic brain under electron microscopy [23], but the molecular pathway linking I/R to autophagy is still poorly understood: NMDA induced cell death in dissociated neuronal cultures activates autophagy via a mechanism that is probably dependent on JNK [38], [39], and in the cortex hypoxia/ischemia is a potent trigger of autophagy, due to the activation of an ER resident translation initiation factor [40].
In order to exclude that the increase in LC3-II expression was caused by a reduction in lysosomal activity or that a defect in autophagosome-lysosome fusion caused vesicular retention in the cytoplasm [41], we evaluated the expression of lysosomal marker. We, showed that the expression of LAMP1, a major constituent of the lysosomal membrane, was increased in damaged GCL-cells from 12 h after I/R, before the finding of LC3 positivity (24 h), but both disappeared at 48 h: this could support the hypothesis that the marked positivity for autophagosome in the GCL-neurons reflect an increase in the autophagic activity more than an inhibition of their clearance. By double-immunofluorescence staining, we confirmed that fusion of autophagosomes with lysosomes occurred in neurons after I/R, underling that the autophagic process was in progress.
LC3-II is the only known protein that specifically associates with autophagosomes [42], whereas the protein LGP120 (LAMP1) is a lysosomal marker; during autophagosome formation LC3-I isoform is converted into LC3-II and then located at both inner and outer membrane. Our antibody cannot discriminate between LC3-I and LC3-II isoforms, but from the subcellular localization of labelling we can infer that it refers to LC3-II: in fact, LC3-I localizes in the cytosol and its immunoreactivity is diffused, whereas LC3-II is membrane-associated and its immunoreactivity is localized in autophagosomes [43].
In addition, we performed LC3 immunoblot analyses and densitometry in control and IOP retina extracts: in fact, the unconjugated (LC3-I) and conjugated (LC3-II) forms can be easily separated by SDS-page. The amount of LC3-II correlates with the number of autophagosomes and immunoblotting of endogenous LC3 represents a relevant method to measure autophagic activity [42]. We found that protein band densities of LC3-I and LC3-II levels in retinas 24 h after IOP were increased compared to control retinas. Taken together, our results indicate that the autophagic flux increases in the retina after IOP.
Clearance of autophagosomes occurs via fusion with lysosomes [44]. In our experiments, the positivity for the lysosomal markers appeared at 12 h. We also studied the relationship between endocytosis and autophagy. These processes are profoundly related to each other, even though endocytosis can occur also in other types of cell death. Clarke and coworkers have documented the occurrence of endocytosis and autophagic cell death in the isthmooptic nucleus of the chick embryo following deafferentation [28], [29] or after blockade of retrograde trophic maintenance from the retina [45]. In the cortex, increased endocytosis precedes cell death [38] and inhibitors of clathrin-mediated endocytosis block excitotoxic cell death in cultured hippocampal neurons [46]. On the other hand, Borsello et al. [30] showed that endocytosis is not a constant feature of all apoptotic neurons and can occur even in the absence of autophagic cell death.
Our findings show that the I/R caused by increased IOP enhances endocytosis of both HRP and FITC labelled dextran into GCL neurons, one day after the insult. We can exclude that the tracers enter the cells across leaky plasma membranes, since the granules are clearly concentrated in round structures of a size comparable to that of endosomes. Moreover, the large size of HRP would prevent it from entering across cell membranes. Activation of autophagy in vivo may represent a protective mechanism used by cells [47]: autophagy genes delay cell death, and the process of autophagy itself may represent a defense mounted by the cell against starvation [48]. Autophagy can also be activated for the purpose of cellular autolysis and self clearance [49], or as a mechanism to remove toxic, multimeric complexes that eventually promote cell death in neurodegenerative diseases [25]. In addition, in many neurodegenerative disorders altered proteins are first degraded through either the ubiquitin-proteasome system, or via chaperone-mediated autophagy, and impairment of these mechanisms promotes protein aggregation [50], [25]. Interestingly, there is a progressive deterioration in autophagic mechanisms with aging [51,52].
On the other hand, autophagy may promote cell death through excessive self digestion, and via the degradation of essential cellular constituents [53]. Several autophagy-related proteins participate in the different steps of autophagy [54]. Autophagic markers such as Beclin-I (Bcl-2-interactin protein) and LC3 (microtubule associated protein 1 light chain 3) are increased in the penumbra of an area of cerebral focal ischemia in the two days following the insult [24]; such increase might represent both a mechanism to recycle damaged material and to lead to cell death.
To verify this controversial hypothesis, we have used a widely used pharmacological inhibitor of autophagy in mammalian cells, 3-MA. 3-MA inhibits the activity of the class III phosphatidylinositol kinase (PI3K), the mammalian homolog of yeast vps34. The latter is required for protein sorting from the Golgi to the lysosome or vacuole in yeast [55]. Therefore 3-MA potently inhibits maturation of autophagosomes and has been commonly used to understand the role of autophagy [56], [57], [58], [59], [60].
The inhibition of autophagy by 3-MA suppressed autophagosome formation and the positivity for apoptotic markers (cleavated caspase-3 and TUNEL staining), thus reducing cell death, underlining that apoptosis and autophagy pathways are intricately intertwined in the cell [20], [61], [62], [63]. By blocking the Moreover, the antiapoptotic protein Bcl-2 is also an antiautophagy protein via its inhibitory interaction with Beclin-I [65]. Beclin-I colocalizes with activated caspase-3 [24] suggesting that autophagy could represent an associated/alternative pathway of cell death which may become particularly relevant when caspasemediated apoptosis is blocked [26]. Another relevant issue regarding autophagic cell death is whether it represents an independent mode of cell death or it occurs in other types of cell death, of which it may represent itself a step [26]. Our results showed that autophagy and apoptosis can coexist in the same damaged neurons after I/R. In fact in different studies, apoptosis and autophagy have been found to share common molecular aspects and they might coexist in the same cell [66]. Nevertheless, we could also find neurons reactive only for either marker, thus suggesting that autophagy and apoptosis do not necessarily overlap and may occur independently from each other. Certainly, autophagic and apoptotic pathways show common upstream signals and their functional relationship is complex [67] and deserves further studies.
Moreover it has been demonstrated that several neurodegenerative diseases elicit the reactive Müller cell gliosis with a rapid and massive GFAP accumulation [68]. Müller cell represents the main retinal glial cell that orchestrates the retinal response after damage [69]. 3-MA treatment prevented astrogliosis in damaged retinas at 24 h, compared to control, demonstrating that inhibition of autophagy partially prevents Müller cells activity during I/R injury.
In conclusion, we have shown not only an enhancement in autophagosome formation but also an increase in lysosomal activity after IOP in the retina, suggesting the improvement of autophagy flux. However, it is unclear whether the enhancement of autophagy activities was due to a defect in lysosomal fusion causing an accumulation of autophagosomes. Nevertheless, the absence of autophagic vesicles at 48 h showed that the process was in progress and they had been degraded.
There is a growing interest in the role of autophagy in neurodegenerative diseases and following ischemia, and an emerging consensus that autophagy represents a double edged  sword, representing alternatively a protective and prosurvival mechanism, or part of a pathway leading to cell death. Targeting autophagy, either by inhibition or by enhancement, could represent a novel and promising tool in the treatment of diseases of the nervous system, in retinal ischemia as well as shown for Alzheimer's and Parkinson's diseases [70] and in a neonatal model of cerebral ischemia [71].

Animals
Adult male Wistar albino rats (4-7 weeks of age) from the animal colony in the Department of Anatomy, Pharmacology and Forensic Medicine at the University of Torino were housed with a 12 h light/ dark cycle, and given free access to food and water. I/R was induced in one group of rats (N = 24) ( Table 1). Four animals of the first group were used for labelling endocytosis, as described below, and three animals were treated with 3-Methyladeninde (3-MA). All efforts were made to minimize the number of animals used.

Induction of I/R by elevated IOP
Induction of I/R was achieved according to the method of Takahashi et al. [7]: rats were anaesthetized by inhalation of 1.5%-3.5% iso fluorane (Forene TM , Deutsche Abbott GmbH, Wiesbaden, Germany) vaporized in a 30/70 mixture of O 2 /N 2 O using a face mask; 0.5% proparacaine hydrochloride solution (an ophthalmic topical anaesthetic, Bausch & Lomb Inc., Tampa-FL) was applied onto the eye. Under the operating microscope (Carl Zeiss Inc., Jena, Germany) a 27-gauge needle, connected to a reservoir containing 500 ml sterile saline, was inserted into the anterior chamber of the left eye. IOP was raised to 110 mmHg by elevating the reservoir 149.6 cm above the animal's eye [72], [73]. The infusion needle was removed from the anterior chamber after one hour, and the IOP was allowed to return to normal levels. The right eye of each rat was cannulated, but was maintained at a normal IOP (28 mmHg) [74] for one hour, as a control. Animals were sacrificed at 12, 24, or 48 h after the increase in IOP.

Acid phosphatase histochemistry
Two histochemical techniques were used for detecting AP activity in the ischemic retina: the technique described by Barka and Anderson [75] was applied for detecting lysosomal enzymes, and the Gomori method was used to study lysosomal membrane activation [76].
For the Gomori procedure, sections were thawed, allowed to dry for 1 hour at 37uC, washed three times in saline, and once in distilled H 2 O (dH 2 O) at room temperature; they were immersed in the incubation medium containing 0.68 mg/ml Pb(NO 3 ) 2 and 4.91 mg/ml sodium glycerophosphate in 0.2 M acetate buffer (pH 4.7) for 90 min at 37uC. Some sections were incubated in medium to which 0.01 M NaF (an AP inhibitor) was added; these served as a negative control (i.e., for no AP activity). The AP reaction was stopped by rinsing slides for 5 min in each of four dH 2 O washes. Endogenous AP activity was revealed by immersing sections in 1:200 ammonium sulphate in dH 2 O for 10-20 seconds, or until a brown precipitate appeared. Sections were rinsed, coverslipped in 1:1 phosphate buffer (PB)-glycerol, and analyzed on a Nikon Eclipse E800 light microscope.
For a modified Barka and Anderson procedure, slides were incubated for 45 minutes at 37uC in a ready-for-use solution containing naphthol AS-BI phosphate, di-methylformamide, pararosanilin in 0.2 M acetate buffer (pH 5.6), and 4% sodium nitrate (Bio-Optica, Milan, Italy); they were washed in dH 2 O for 10 minutes, counterstained with buffered methyl green (Bio-Optica) for 5 minutes, and coverslipped in 1:1 phosphate-buffered saline (PBS)-glycerol. This method reveals enzymatic activity by a red precipitate, while nuclei are labelled green.
Control sections to verify the specificity of the secondary antibodies was reacted similarly, except the primary antibody was omitted in incubation. No immunolabeling was seen in control sections (data not shown).

Immunoblotting
Animals were sacrificed 24 h after the increase in IOP, retinas were dissected from the sclera and then immediately homogenized in a glass-Teflon Potter homogenizer in an ice-cold lysis buffer containing 20 mM Hepes, pH 7.5/10 mM KCl/1.5 mM MgCl2/1 mM ethylenediaminetetraacetic acid (EDTA)/1 mM ethylene glycol tetraacetic acid (EGTA)/1 mM DTT/0.5% CHAPS/complete protease inhibitors; Roche Cat. No. 11 697 498 001) the homogenates were centrifuged at 12000 rpm for 15 min at 4uC. Protein concentration was determined using a Bradford assay #23236. Proteins extracts (50 mg) were separated on SDS-PAGE (12% polyacrilamide) and transferred to polyvinylidene difluoride membranes. Then the membranes were blocked in 5% nonfat milk in tris buffered saline (TBS)-T (200 mM Tris and 1.5 M NaCl with 0.1% Tween 20) and were incubated with primary antibody diluted in TBS-T overnight at 4uC.
The membranes were washed and incubated with secondary horseradish peroxidase-coupled antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) in TBS-T for 1 hour at room temperature. After the final washes, the proteins were detected by enhanced chemiluminescence. The bands were quantified using Quantity OneH 1-D Analysis Software (Bio-Rad Laboratories) and values were normalized with respect to tubulin. The values were then expressed as a percentage relative to the sham level of OD. The antibodies used were as follows: anti-LC3 (L8918) from Sigma-Aldrich (1:1000), anti-b III tubulin (1637) from Millipore (1:400), goat anti-mouse IgG-HRP (sc-2005) from Santa Cruz Biotechnology (1:10000) and goat anti-rabbit IgG-HRP (sc-2004) from Santa Cruz Biotechnology (1:10000).

Labeling endocytosis in vivo
Endocytosis was documented by using one of two tracers: 10% horseradish peroxidase (HRP, Sigma-Aldrich, type VI-A) in saline, or 25 mg/ml 4.4 kDa fluorescein isothiocyanate (FITC)-labelled dextran (Sigma-Aldrich, St. Louis, MO) in saline. Each tracer was injected into the I/R eye and the control eye of two rats after the IOP was elevated; after three hours, the animals were killed, perfused, and retinas were collected and sectioned, as above, for histology. The two tracers were selected for the difference in their sizes: HRP (44 kDa) is a much larger molecule than dextran (4.4 kDa). To reveal HRP, sections were incubated with 10X 3,3 diaminobendizine/metal concentrate diluted to 1X in peroxide buffer (Roche Diagnostics GmbH, Mannheim, Germany) for 5 minutes, rinsed in dH 2 O for 5 minutes, counterstained with methyl green to label cell nuclei and coverslipped in PBS-glycerol (1:1).Sections from animals that had received injections of FITClabelled dextran were analyzed on a Nikon Eclipse E800 epifluorescence microscope, with the use of appropriate filters to document FITC-positive vacuoles; bisbenzimide (Hoechst # 33258, Sigma, St.Louis, MO) was used to counterstain cell nuclei.

Cell counts
Nuclear profiles were counted under a 40x objective in five nonadjacent 100 mm-spaced sections from each retina all along the ganglion cell layer (GCL), starting at a distance 100 mm from the optic nerve head to the ora serrata. Nuclear profile counts (number of cells/500 mm) are expressed as mean among the animals6standard deviation (SD). The raw number of neurons in the GCL was corrected using the Abercrombie's correction factor [77], [78]. The statistical significance of the results was determined by hieralchical ANOVA followed by planned pair-wise comparisons with the Tukey test. P-values ,0.05 are considered significant.

3-MA administration
3-Methyladeninde (Sigma-Aldrich, St. Louis, MO) was diluited in saline (0.9% NaCl) to a stock solution of 100 mM and vitreal injection was performed in the I/R retina (n = 3) with 3 ml (<33 mM). Rats were injected at the end of I/R injury, under anaesthesia.